A unique, human-operated chatbot initiative named Quili.AI was launched by approximately 50 residents of a community outside Chile’s capital, Santiago. This 12-hour project, which took place on a Saturday, aimed to highlight the significant environmental toll, particularly the water footprint, associated with artificial intelligence data centers in the region. The organizers reported that the system successfully fielded more than 25,000 requests from around the globe.

Unlike conventional AI chatbots such as ChatGPT or Google’s Gemini, Quili.AI deliberately provided delayed, human-generated responses. For instance, a request to generate an image of a “sloth playing in the snow” did not produce an instant output. Instead, a human volunteer responded in Spanish, asking the user to wait a few moments and reminding them that a person was handling the request. About 10 minutes later, a hand-drawn, penciled sketch of a cartoonish sloth clutching a snowball appeared.

Lorena Antiman of the environmental group Corporación NGEN, an organizer of the event, articulated the project's core objective: “The goal is to highlight the hidden water footprint behind AI prompting and encourage more responsible use.” The responses were crafted by a rotating team of volunteers working on laptops within a community center in Quilicura. This municipality, situated at the urban periphery of Santiago, has become a burgeoning hub for data centers. When an Associated Press reporter inquired about the identity of the sloth drawing's creator, the website revealed it was a local youth assisting with illustrations.

Quili.AI’s approach underscored the value of human knowledge and deliberation. Antiman explained, “Quili.AI isn’t about always having an instant answer. It’s about recognizing that not every question needs one.” Volunteers were empowered to openly state when they didn't know an answer, share their perspective, or respond with curiosity rather than striving for immediate certainty. This initiative is not designed to reject the “incredibly valuable” uses of AI but rather to foster critical thinking about the impacts of extensive “casual prompting” in water-stressed regions like Quilicura.

The campaign emerged amidst a growing debate, both in Chile and internationally, concerning the substantial environmental costs associated with AI usage. Data center computer chips, which are essential for running AI systems, demand immense amounts of electricity. Furthermore, some facilities require large volumes of water for cooling purposes, with usage levels varying depending on their geographical location and the specific type of equipment employed. Major cloud computing entities, including Amazon, Google, and Microsoft, are among several companies that have either constructed or planned data centers within the Santiago region.

Google has asserted that its Quilicura data center, activated in 2015, stands as the “most energy efficient in Latin America” and has emphasized its investments in wetlands restoration and irrigation projects within the surrounding Maipo River basin. However, the company has also encountered a court challenge regarding another project near Santiago, stemming from concerns over water usage. This context is particularly pertinent given that Chile has endured a decade of severe drought, a factor that experts believe contributed to the spread of recent deadly wildfires, further highlighting the urgency of water conservation efforts.